Exploratory Plots for 2017-2018 Acoustic/Fish Data
Purpose To explore the Acoustic data gathered in 2017 and 2018 to expose important trends between sites, diurnal patterns, fish abundance, lunar phase, and coral reef acoustics.
Combined Model All variables are matched to the files that were used for Fish call counts (3:00, 9:00, 15:00, 21:00)
Red is the Mean Line
Blue is the Median Line
## `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.
## `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.
Stats for Mid Frequency SPL
## Min. 1st Qu. Median Mean 3rd Qu. Max.
## 100.2 103.4 104.7 105.8 107.7 119.3
Variance of MF
## [1] 11.87399
Stats for High Frequency SPL
## Min. 1st Qu. Median Mean 3rd Qu. Max.
## 100.4 114.5 117.2 117.1 119.2 129.1
Variance of HF
## [1] 11.96909
ACI histogram
## `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.
Running basic regressions linking the explanatory to the response at their lowest levels and combined to see how different sites/ hours change the regression - SPL
Linear Model outputs below each
##
## Call:
## lm(formula = SPL_HF ~ Snaps, data = Snap.HF17)
##
## Residuals:
## Min 1Q Median 3Q Max
## -7.8309 -1.9842 0.2062 1.8451 13.3944
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 1.053e+02 6.541e-01 160.99 <2e-16 ***
## Snaps 7.227e-03 4.475e-04 16.15 <2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 2.807 on 10163 degrees of freedom
## Multiple R-squared: 0.02502, Adjusted R-squared: 0.02493
## F-statistic: 260.8 on 1 and 10163 DF, p-value: < 2.2e-16
2017 Snap data, snaps significant.
When you break it down by site, site 32 has the opposite relationship with high frequency and snaps.
2017 Snap/HF SPL Site Breakdown
##
## Call:
## lm(formula = SPL_HF ~ Snaps, data = s17s5)
##
## Residuals:
## Min 1Q Median 3Q Max
## -6.0817 -2.1540 0.4371 1.9805 7.0937
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 87.830664 1.873329 46.88 <2e-16 ***
## Snaps 0.018381 0.001277 14.39 <2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 2.483 on 2101 degrees of freedom
## Multiple R-squared: 0.08971, Adjusted R-squared: 0.08928
## F-statistic: 207.1 on 1 and 2101 DF, p-value: < 2.2e-16
##
## Call:
## lm(formula = SPL_HF ~ Snaps, data = s17s8)
##
## Residuals:
## Min 1Q Median 3Q Max
## -6.3374 -1.3945 0.1363 1.4230 9.4265
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 7.185e+01 1.270e+00 56.59 <2e-16 ***
## Snaps 3.314e-02 9.084e-04 36.48 <2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 2.117 on 1831 degrees of freedom
## Multiple R-squared: 0.4209, Adjusted R-squared: 0.4206
## F-statistic: 1331 on 1 and 1831 DF, p-value: < 2.2e-16
##
## Call:
## lm(formula = SPL_HF ~ Snaps, data = s17s35)
##
## Residuals:
## Min 1Q Median 3Q Max
## -6.9213 -1.7565 -0.0424 1.6512 10.3407
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 71.282701 1.451690 49.10 <2e-16 ***
## Snaps 0.029598 0.000995 29.75 <2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 2.573 on 2205 degrees of freedom
## Multiple R-squared: 0.2864, Adjusted R-squared: 0.2861
## F-statistic: 884.9 on 1 and 2205 DF, p-value: < 2.2e-16
##
## Call:
## lm(formula = SPL_HF ~ Snaps, data = s17s40)
##
## Residuals:
## Min 1Q Median 3Q Max
## -5.1902 -1.2312 0.0344 1.2186 9.3897
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 7.644e+01 1.044e+00 73.19 <2e-16 ***
## Snaps 2.679e-02 7.062e-04 37.93 <2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 1.736 on 1862 degrees of freedom
## Multiple R-squared: 0.4359, Adjusted R-squared: 0.4356
## F-statistic: 1439 on 1 and 1862 DF, p-value: < 2.2e-16
##
## Call:
## lm(formula = SPL_HF ~ Snaps, data = s17s32)
##
## Residuals:
## Min 1Q Median 3Q Max
## -4.936 -1.084 0.114 1.063 7.102
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 137.43721 0.89844 152.97 <2e-16 ***
## Snaps -0.01414 0.00060 -23.56 <2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 1.532 on 2156 degrees of freedom
## Multiple R-squared: 0.2047, Adjusted R-squared: 0.2044
## F-statistic: 555 on 1 and 2156 DF, p-value: < 2.2e-16
2018 Snap/HF SPL
Removing outliers
##
## Call:
## lm(formula = SPL_HF ~ Snaps, data = Snap.HF18)
##
## Residuals:
## Min 1Q Median 3Q Max
## -8.6746 -2.0071 -0.0087 2.3005 12.6859
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 81.468599 1.919126 42.45 <2e-16 ***
## Snaps 0.025921 0.001315 19.71 <2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 3.072 on 1453 degrees of freedom
## Multiple R-squared: 0.211, Adjusted R-squared: 0.2105
## F-statistic: 388.7 on 1 and 1453 DF, p-value: < 2.2e-16
2018 Snap data with outliers removed. Snaps significant.
When split by sight, site 32 has a flat relationship.
2018 Snap/HF SPL Site Breakdown
##
## Call:
## lm(formula = SPL_HF ~ Snaps, data = s18s5)
##
## Residuals:
## Min 1Q Median 3Q Max
## -5.9245 -1.5844 0.1253 1.6517 5.3652
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 60.225493 4.141984 14.54 <2e-16 ***
## Snaps 0.038216 0.002823 13.54 <2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 2.238 on 289 degrees of freedom
## Multiple R-squared: 0.388, Adjusted R-squared: 0.3859
## F-statistic: 183.2 on 1 and 289 DF, p-value: < 2.2e-16
##
## Call:
## lm(formula = SPL_HF ~ Snaps, data = s18s8)
##
## Residuals:
## Min 1Q Median 3Q Max
## -4.7323 -1.3490 -0.0334 1.4302 4.1632
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 68.048894 2.745679 24.78 <2e-16 ***
## Snaps 0.035631 0.001889 18.87 <2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 1.849 on 289 degrees of freedom
## Multiple R-squared: 0.5519, Adjusted R-squared: 0.5504
## F-statistic: 356 on 1 and 289 DF, p-value: < 2.2e-16
##
## Call:
## lm(formula = SPL_HF ~ Snaps, data = s18s35)
##
## Residuals:
## Min 1Q Median 3Q Max
## -3.9232 -1.1784 -0.1059 1.0416 7.5440
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 83.11812 1.96366 42.33 <2e-16 ***
## Snaps 0.02652 0.00133 19.94 <2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 1.739 on 289 degrees of freedom
## Multiple R-squared: 0.5791, Adjusted R-squared: 0.5776
## F-statistic: 397.6 on 1 and 289 DF, p-value: < 2.2e-16
##
## Call:
## lm(formula = SPL_HF ~ Snaps, data = s18s40)
##
## Residuals:
## Min 1Q Median 3Q Max
## -4.4914 -1.3764 -0.1106 1.2747 6.8204
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 68.686141 2.640335 26.01 <2e-16 ***
## Snaps 0.033362 0.001816 18.38 <2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 1.869 on 289 degrees of freedom
## Multiple R-squared: 0.5388, Adjusted R-squared: 0.5372
## F-statistic: 337.7 on 1 and 289 DF, p-value: < 2.2e-16
##
## Call:
## lm(formula = SPL_HF ~ Snaps, data = s18s32)
##
## Residuals:
## Min 1Q Median 3Q Max
## -3.6331 -1.9735 0.2795 1.8090 4.4940
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 1.213e+02 3.090e+00 39.250 <2e-16 ***
## Snaps -2.742e-04 2.136e-03 -0.128 0.898
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 2.074 on 289 degrees of freedom
## Multiple R-squared: 5.699e-05, Adjusted R-squared: -0.003403
## F-statistic: 0.01647 on 1 and 289 DF, p-value: 0.898
Mid Frequency - SPL
##
## Call:
## lm(formula = SPL_Midrange ~ Tot_Knocks, data = AC.DF1)
##
## Residuals:
## Min 1Q Median 3Q Max
## -7.2088 -2.1924 -0.7869 1.6962 11.9299
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 1.045e+02 3.773e-01 277.047 < 2e-16 ***
## Tot_Knocks 1.801e-02 4.252e-03 4.237 3.54e-05 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 3.302 on 189 degrees of freedom
## Multiple R-squared: 0.08674, Adjusted R-squared: 0.08191
## F-statistic: 17.95 on 1 and 189 DF, p-value: 3.538e-05
Mid Frequency - ACI
##
## Call:
## glm(formula = ACI_Midrange ~ Tot_Knocks + Year, family = "Gamma",
## data = AC.DF1)
##
## Deviance Residuals:
## Min 1Q Median 3Q Max
## -0.24710 -0.14537 -0.07297 0.11790 0.40777
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 1.746e-05 3.758e-07 46.459 <2e-16 ***
## Tot_Knocks -8.617e-09 3.462e-09 -2.489 0.0137 *
## Year18 2.779e-07 4.078e-07 0.682 0.4964
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for Gamma family taken to be 0.02745441)
##
## Null deviance: 4.9867 on 190 degrees of freedom
## Residual deviance: 4.8131 on 188 degrees of freedom
## AIC: 4038.2
##
## Number of Fisher Scoring iterations: 4
##
## Call:
## glm(formula = ACI_Midrange ~ Num_L_calls + Year, family = "Gamma",
## data = AC.DF1)
##
## Deviance Residuals:
## Min 1Q Median 3Q Max
## -0.19561 -0.15304 -0.06392 0.11642 0.42225
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 1.690e-05 3.382e-07 49.964 <2e-16 ***
## Num_L_calls -4.017e-09 2.953e-08 -0.136 0.892
## Year18 2.304e-07 4.140e-07 0.557 0.579
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for Gamma family taken to be 0.02823657)
##
## Null deviance: 4.9867 on 190 degrees of freedom
## Residual deviance: 4.9772 on 188 degrees of freedom
## AIC: 4044.6
##
## Number of Fisher Scoring iterations: 4
##
## Call:
## glm(formula = ACI_Midrange ~ Num_Herbivory + Year, family = "Gamma",
## data = AC.DF1)
##
## Deviance Residuals:
## Min 1Q Median 3Q Max
## -0.22182 -0.15000 -0.06905 0.11317 0.42714
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 1.697e-05 3.019e-07 56.198 <2e-16 ***
## Num_Herbivory -2.940e-08 2.290e-08 -1.284 0.201
## Year18 2.350e-07 4.112e-07 0.571 0.568
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for Gamma family taken to be 0.02795665)
##
## Null deviance: 4.9867 on 190 degrees of freedom
## Residual deviance: 4.9338 on 188 degrees of freedom
## AIC: 4042.9
##
## Number of Fisher Scoring iterations: 4
##
## Call:
## glm(formula = ACI_Midrange ~ Tot_Knocks + Site + Year + Hour,
## family = "Gamma", data = AC.DF1)
##
## Deviance Residuals:
## Min 1Q Median 3Q Max
## -0.23464 -0.14810 -0.02786 0.09402 0.40045
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 1.687e-05 5.893e-07 28.620 < 2e-16 ***
## Tot_Knocks -6.857e-09 4.115e-09 -1.666 0.09738 .
## Site35 -3.948e-07 6.308e-07 -0.626 0.53219
## Site40 1.842e-06 6.744e-07 2.731 0.00694 **
## Site5 -1.827e-07 7.031e-07 -0.260 0.79525
## Site8 -2.771e-07 6.134e-07 -0.452 0.65206
## Year18 2.542e-07 3.925e-07 0.648 0.51800
## Hour21 6.213e-07 5.771e-07 1.077 0.28311
## Hour3 6.336e-07 5.689e-07 1.114 0.26689
## Hour9 1.735e-07 5.654e-07 0.307 0.75935
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for Gamma family taken to be 0.02532824)
##
## Null deviance: 4.9867 on 190 degrees of freedom
## Residual deviance: 4.3779 on 181 degrees of freedom
## AIC: 4034
##
## Number of Fisher Scoring iterations: 4
Breakdown by Site - SPL
Breakdown by Site - ACI
Breakdown by Hour - SPL
Breakdown by Hour - ACI
3 AM, long calls don’t seem to explain a great deal of the relationship at any site
9 AM, long calls don’t seem to explain the relationship at any site
3 PM, long calls don’t seem to explain the relationship
9 PM, long calls don’t seem to explain the relationship
3 AM, Extremely low herbivory at all sites. No relationship
Again, extremely low herbivory, no relationship.
Higher herbivory. Seems like there is a relationship at site 40, 8, and 35.
Higher herbivory here as well, although there is no positive relationship at any site.
Summary Knocks significantly explained SPLMF at sites 35 and 32 and at 9AM.
Acoustics Breakdown All acoustic metrics (SPL and ACI) are broken down into 2 frequency bands: High Frequency (Frequencies between 1 kHz - 22 kHz) and Mid Frequency (Frequencies between 160 Hz and 1 kHz)
Note 2017 had a 10 minute duty cycle with 5 minutes recording while 2018 had a 15 minute duty cycle with 5 minutes recording, so the number of files averages differs between years
Total Deployment Plots
Preliminary Models Looking into the relationships between biogenic sounds (Knocks/Calls and Snaps) and their frequency spectra (MF SPL/HF SPL) respectively.
shapiro.test(AC.DF1$SPL_Midrange)
##
## Shapiro-Wilk normality test
##
## data: AC.DF1$SPL_Midrange
## W = 0.92679, p-value = 3.406e-08
qqnorm(AC.DF1$SPL_Midrange)
#ks.test(SPLHF.long$SPL_HF, "pnorm", mean=mean(SPLHF.long$SPL_HF), sd=sd(SPLHF.long$SPL_HF))
#ks.test(SPLMF.long$SPL_MF, "pnorm", mean=mean(SPLMF.long$SPL_MF), sd=sd(SPLMF.long$SPL_MF))
ggplot(data = Snap.HF, aes(Snap.HF$SPL_HF)) + geom_histogram() + ggtitle("HF SPL distribution")
## `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.
ggplot(data = AC.DF1, aes(AC.DF1$ACI_Midrange)) + geom_histogram() + ggtitle("MF ACI distribution")
## `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.
ggplot(data = AC.DF1, aes(AC.DF1$ACI_HF)) + geom_histogram() + ggtitle("HF ACI distribution")
## `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.
gamma_test(AC.DF1$ACI_Midrange)
##
## Test of fit for the Gamma distribution
##
## data: AC.DF1$ACI_Midrange
## V = 4.3336, p-value = 0.002182
gamma_test(AC.DF1$ACI_HF)
##
## Test of fit for the Gamma distribution
##
## data: AC.DF1$ACI_HF
## V = 4.9011, p-value = 0.0005291
gamma_test(Snap.HF$SPL_HF)
##
## Test of fit for the Gamma distribution
##
## data: Snap.HF$SPL_HF
## V = 7.8948, p-value = 2.372e-08
Don’t seem to have a normal distribution here… Working on testing different distributions. Can’t find what the p-values indicate for these gamma tests
Maximal Model with Bill
fit.m <- lm(SPL_Midrange ~(Tot_Knocks + Num_Herbivory + Num_L_calls)*(Site + Hour) + Year, data = AC.DF1Co)
summary(fit.m)
##
## Call:
## lm(formula = SPL_Midrange ~ (Tot_Knocks + Num_Herbivory + Num_L_calls) *
## (Site + Hour) + Year, data = AC.DF1Co)
##
## Residuals:
## Min 1Q Median 3Q Max
## -4.2170 -1.2622 -0.1569 1.1008 5.9507
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 1.040e+02 1.039e+00 100.154 < 2e-16 ***
## Tot_Knocks 6.742e-03 1.513e-02 0.446 0.656421
## Num_Herbivory 2.019e-01 3.082e-01 0.655 0.513320
## Num_L_calls -4.731e-02 1.181e-01 -0.401 0.689255
## Site35 9.601e-01 1.166e+00 0.823 0.411629
## Site40 -2.755e+00 1.070e+00 -2.574 0.010960 *
## Site5 -5.481e-01 1.095e+00 -0.501 0.617374
## Site8 -1.808e+00 1.038e+00 -1.743 0.083340 .
## Hour21 8.820e-01 5.437e-01 1.622 0.106767
## Hour3 -2.264e+00 2.092e+00 -1.082 0.280983
## Hour9 2.618e+00 2.096e+00 1.249 0.213552
## Year18 3.925e+00 3.196e-01 12.282 < 2e-16 ***
## Tot_Knocks:Site35 -1.458e-02 1.432e-02 -1.018 0.310289
## Tot_Knocks:Site40 1.099e-02 1.726e-02 0.636 0.525394
## Tot_Knocks:Site5 -1.831e-02 1.408e-02 -1.301 0.195163
## Tot_Knocks:Site8 -1.516e-02 1.522e-02 -0.996 0.320693
## Tot_Knocks:Hour21 8.699e-03 1.092e-02 0.797 0.426879
## Tot_Knocks:Hour3 1.184e-02 1.102e-02 1.074 0.284454
## Tot_Knocks:Hour9 4.361e-02 1.171e-02 3.723 0.000273 ***
## Num_Herbivory:Site35 -1.447e-01 3.092e-01 -0.468 0.640479
## Num_Herbivory:Site40 -5.393e-01 3.150e-01 -1.712 0.088813 .
## Num_Herbivory:Site5 -2.400e-01 3.076e-01 -0.780 0.436461
## Num_Herbivory:Site8 -7.112e-02 3.085e-01 -0.231 0.817956
## Num_Herbivory:Hour21 -6.096e-04 8.470e-02 -0.007 0.994267
## Num_Herbivory:Hour3 -7.106e-01 6.836e-01 -1.040 0.300117
## Num_Herbivory:Hour9 5.260e-01 7.094e-01 0.741 0.459511
## Num_L_calls:Site35 2.410e-01 1.910e-01 1.262 0.208822
## Num_L_calls:Site40 1.828e-01 1.078e-01 1.696 0.091819 .
## Num_L_calls:Site5 -7.132e-02 1.396e-01 -0.511 0.610178
## Num_L_calls:Site8 4.688e-02 1.028e-01 0.456 0.648863
## Num_L_calls:Hour21 8.163e-02 9.719e-02 0.840 0.402248
## Num_L_calls:Hour3 3.101e-02 1.165e-01 0.266 0.790544
## Num_L_calls:Hour9 -2.675e-01 1.351e-01 -1.980 0.049427 *
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 2.041 on 158 degrees of freedom
## Multiple R-squared: 0.7084, Adjusted R-squared: 0.6493
## F-statistic: 12 on 32 and 158 DF, p-value: < 2.2e-16
stepAIC(fit.m)
## Start: AIC=302.21
## SPL_Midrange ~ (Tot_Knocks + Num_Herbivory + Num_L_calls) * (Site +
## Hour) + Year
##
## Df Sum of Sq RSS AIC
## - Num_Herbivory:Hour 3 6.83 664.69 298.19
## - Tot_Knocks:Site 4 24.83 682.68 301.29
## <none> 657.86 302.21
## - Num_L_calls:Site 4 36.35 694.20 304.48
## - Num_L_calls:Hour 3 41.50 699.35 307.90
## - Num_Herbivory:Site 4 79.04 736.90 315.88
## - Tot_Knocks:Hour 3 92.30 750.16 321.29
## - Year 1 628.11 1285.96 428.24
##
## Step: AIC=298.19
## SPL_Midrange ~ Tot_Knocks + Num_Herbivory + Num_L_calls + Site +
## Hour + Year + Tot_Knocks:Site + Tot_Knocks:Hour + Num_Herbivory:Site +
## Num_L_calls:Site + Num_L_calls:Hour
##
## Df Sum of Sq RSS AIC
## - Tot_Knocks:Site 4 24.85 689.54 297.20
## <none> 664.69 298.19
## - Num_L_calls:Site 4 35.64 700.33 300.16
## - Num_L_calls:Hour 3 40.27 704.96 303.42
## - Tot_Knocks:Hour 3 88.49 753.18 316.06
## - Num_Herbivory:Site 4 96.88 761.56 316.17
## - Year 1 647.13 1311.81 426.04
##
## Step: AIC=297.2
## SPL_Midrange ~ Tot_Knocks + Num_Herbivory + Num_L_calls + Site +
## Hour + Year + Tot_Knocks:Hour + Num_Herbivory:Site + Num_L_calls:Site +
## Num_L_calls:Hour
##
## Df Sum of Sq RSS AIC
## <none> 689.54 297.20
## - Num_L_calls:Site 4 36.89 726.43 299.15
## - Num_L_calls:Hour 3 50.99 740.52 304.82
## - Num_Herbivory:Site 4 101.98 791.52 315.54
## - Tot_Knocks:Hour 3 101.17 790.70 317.34
## - Year 1 655.66 1345.20 422.84
##
## Call:
## lm(formula = SPL_Midrange ~ Tot_Knocks + Num_Herbivory + Num_L_calls +
## Site + Hour + Year + Tot_Knocks:Hour + Num_Herbivory:Site +
## Num_L_calls:Site + Num_L_calls:Hour, data = AC.DF1Co)
##
## Coefficients:
## (Intercept) Tot_Knocks Num_Herbivory
## 103.883544 -0.004243 0.281592
## Num_L_calls Site35 Site40
## -0.026390 1.211723 -3.008499
## Site5 Site8 Hour21
## -0.733301 -1.642642 0.779976
## Hour3 Hour9 Year18
## -0.199326 1.006530 3.904610
## Tot_Knocks:Hour21 Tot_Knocks:Hour3 Tot_Knocks:Hour9
## 0.009165 0.009517 0.040037
## Num_Herbivory:Site35 Num_Herbivory:Site40 Num_Herbivory:Site5
## -0.223915 -0.607692 -0.328665
## Num_Herbivory:Site8 Num_L_calls:Site35 Num_L_calls:Site40
## -0.148352 0.223663 0.137930
## Num_L_calls:Site5 Num_L_calls:Site8 Num_L_calls:Hour21
## -0.132447 0.004349 0.109991
## Num_L_calls:Hour3 Num_L_calls:Hour9
## 0.050565 -0.257516
Next is the best model from AIC stepwise model selection
#SPL_Midrange ~ Tot_Knocks + Num_Herbivory + Num_L_calls + Site + Hour + Year + Tot_Knocks:Site + Tot_Knocks:Hour + Num_Herbivory:Site + Num_L_calls:Site + Num_L_calls:Hour
fit.m2 <- lm(SPL_Midrange ~ Tot_Knocks + Num_Herbivory + Num_L_calls + Site + Hour + Year + Tot_Knocks:Site + Tot_Knocks:Hour + Num_Herbivory:Site + Num_L_calls:Site + Num_L_calls:Hour, data = AC.DF1Co)
Anova(fit.m2, type=3)
## Anova Table (Type III tests)
##
## Response: SPL_Midrange
## Sum Sq Df F value Pr(>F)
## (Intercept) 44532 1 10786.5010 < 2.2e-16 ***
## Tot_Knocks 1 1 0.1624 0.6874574
## Num_Herbivory 4 1 0.9933 0.3204412
## Num_L_calls 1 1 0.1826 0.6697040
## Site 120 4 7.2699 2.087e-05 ***
## Hour 42 3 3.3839 0.0196605 *
## Year 647 1 156.7463 < 2.2e-16 ***
## Tot_Knocks:Site 25 4 1.5048 0.2032243
## Tot_Knocks:Hour 88 3 7.1449 0.0001554 ***
## Num_Herbivory:Site 97 4 5.8663 0.0001968 ***
## Num_L_calls:Site 36 4 2.1582 0.0760730 .
## Num_L_calls:Hour 40 3 3.2514 0.0233376 *
## Residuals 665 161
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
summary(fit.m2)
##
## Call:
## lm(formula = SPL_Midrange ~ Tot_Knocks + Num_Herbivory + Num_L_calls +
## Site + Hour + Year + Tot_Knocks:Site + Tot_Knocks:Hour +
## Num_Herbivory:Site + Num_L_calls:Site + Num_L_calls:Hour,
## data = AC.DF1Co)
##
## Residuals:
## Min 1Q Median 3Q Max
## -4.2174 -1.2322 -0.1251 1.0980 5.9555
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 104.206794 1.003358 103.858 < 2e-16 ***
## Tot_Knocks 0.005885 0.014603 0.403 0.687457
## Num_Herbivory 0.278351 0.279293 0.997 0.320441
## Num_L_calls -0.050139 0.117328 -0.427 0.669704
## Site35 0.766076 1.120805 0.684 0.495270
## Site40 -2.936460 1.041347 -2.820 0.005408 **
## Site5 -0.742531 1.046012 -0.710 0.478813
## Site8 -2.085095 0.990076 -2.106 0.036756 *
## Hour21 0.879295 0.538575 1.633 0.104501
## Hour3 -0.168172 0.562300 -0.299 0.765265
## Hour9 1.120116 0.608855 1.840 0.067653 .
## Year18 3.942663 0.314913 12.520 < 2e-16 ***
## Tot_Knocks:Site35 -0.013232 0.014091 -0.939 0.349122
## Tot_Knocks:Site40 0.011545 0.017084 0.676 0.500151
## Tot_Knocks:Site5 -0.017482 0.013708 -1.275 0.204032
## Tot_Knocks:Site8 -0.015266 0.014960 -1.020 0.309036
## Tot_Knocks:Hour21 0.008873 0.010204 0.869 0.385874
## Tot_Knocks:Hour3 0.012101 0.010872 1.113 0.267341
## Tot_Knocks:Hour9 0.042694 0.011519 3.706 0.000289 ***
## Num_Herbivory:Site35 -0.221091 0.281598 -0.785 0.433530
## Num_Herbivory:Site40 -0.611817 0.298717 -2.048 0.042169 *
## Num_Herbivory:Site5 -0.317416 0.282805 -1.122 0.263369
## Num_Herbivory:Site8 -0.147308 0.280869 -0.524 0.600671
## Num_L_calls:Site35 0.247525 0.189782 1.304 0.194006
## Num_L_calls:Site40 0.183437 0.106859 1.717 0.087970 .
## Num_L_calls:Site5 -0.069899 0.138445 -0.505 0.614327
## Num_L_calls:Site8 0.053260 0.101700 0.524 0.601206
## Num_L_calls:Hour21 0.080441 0.096709 0.832 0.406763
## Num_L_calls:Hour3 0.044625 0.115290 0.387 0.699218
## Num_L_calls:Hour9 -0.258808 0.133308 -1.941 0.053953 .
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 2.032 on 161 degrees of freedom
## Multiple R-squared: 0.7054, Adjusted R-squared: 0.6523
## F-statistic: 13.29 on 29 and 161 DF, p-value: < 2.2e-16
plot(fit.m2)
## NULL
## NULL
Maximal model following Bill’s method
fit.a <- glm(ACI_Midrange ~(Tot_Knocks + Num_Herbivory + Num_L_calls)*(Site + Hour) + Year, data = AC.DF1Co, family = "Gamma")
summary(fit.a)
##
## Call:
## glm(formula = ACI_Midrange ~ (Tot_Knocks + Num_Herbivory + Num_L_calls) *
## (Site + Hour) + Year, family = "Gamma", data = AC.DF1Co)
##
## Deviance Residuals:
## Min 1Q Median 3Q Max
## -0.24879 -0.12333 -0.02497 0.08025 0.36994
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 1.690e-05 1.444e-06 11.699 <2e-16 ***
## Tot_Knocks -2.546e-08 1.928e-08 -1.321 0.188
## Num_Herbivory 3.467e-07 4.468e-07 0.776 0.439
## Num_L_calls -3.403e-08 1.565e-07 -0.217 0.828
## Site35 -1.184e-06 1.608e-06 -0.736 0.463
## Site40 1.624e-06 1.531e-06 1.061 0.290
## Site5 -1.438e-06 1.512e-06 -0.952 0.343
## Site8 -5.345e-07 1.453e-06 -0.368 0.713
## Hour21 6.913e-07 7.114e-07 0.972 0.333
## Hour3 -1.930e-06 2.549e-06 -0.757 0.450
## Hour9 3.315e-06 3.230e-06 1.026 0.306
## Year18 3.223e-07 4.236e-07 0.761 0.448
## Tot_Knocks:Site35 1.090e-08 1.827e-08 0.597 0.551
## Tot_Knocks:Site40 2.652e-08 2.402e-08 1.104 0.271
## Tot_Knocks:Site5 2.536e-08 1.802e-08 1.407 0.161
## Tot_Knocks:Site8 2.288e-08 1.972e-08 1.160 0.248
## Tot_Knocks:Hour21 6.822e-09 1.391e-08 0.490 0.625
## Tot_Knocks:Hour3 -8.492e-10 1.403e-08 -0.061 0.952
## Tot_Knocks:Hour9 6.244e-09 1.483e-08 0.421 0.674
## Num_Herbivory:Site35 -2.623e-07 4.484e-07 -0.585 0.559
## Num_Herbivory:Site40 -3.463e-07 4.595e-07 -0.754 0.452
## Num_Herbivory:Site5 -3.652e-07 4.459e-07 -0.819 0.414
## Num_Herbivory:Site8 -3.946e-07 4.468e-07 -0.883 0.379
## Num_Herbivory:Hour21 -1.418e-07 1.063e-07 -1.335 0.184
## Num_Herbivory:Hour3 -8.192e-07 8.323e-07 -0.984 0.326
## Num_Herbivory:Hour9 1.189e-06 1.088e-06 1.093 0.276
## Num_L_calls:Site35 1.946e-08 2.478e-07 0.079 0.938
## Num_L_calls:Site40 6.352e-08 1.454e-07 0.437 0.663
## Num_L_calls:Site5 -4.761e-08 1.803e-07 -0.264 0.792
## Num_L_calls:Site8 1.097e-09 1.343e-07 0.008 0.993
## Num_L_calls:Hour21 3.364e-09 1.307e-07 0.026 0.979
## Num_L_calls:Hour3 -1.778e-07 1.564e-07 -1.137 0.257
## Num_L_calls:Hour9 -1.595e-07 1.741e-07 -0.916 0.361
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for Gamma family taken to be 0.02551326)
##
## Null deviance: 4.9867 on 190 degrees of freedom
## Residual deviance: 3.8459 on 158 degrees of freedom
## AIC: 4055.2
##
## Number of Fisher Scoring iterations: 4
stepAIC(fit.a)
## Start: AIC=4055.17
## ACI_Midrange ~ (Tot_Knocks + Num_Herbivory + Num_L_calls) * (Site +
## Hour) + Year
##
## Df Deviance AIC
## - Num_L_calls:Site 4 3.8643 4047.9
## - Tot_Knocks:Hour 3 3.8658 4049.9
## - Tot_Knocks:Site 4 3.9294 4050.4
## - Num_L_calls:Hour 3 3.9306 4052.5
## - Num_Herbivory:Site 4 3.9975 4053.1
## - Num_Herbivory:Hour 3 3.9531 4053.4
## - Year 1 3.8606 4053.7
## <none> 3.8459 4055.2
##
## Step: AIC=4048.08
## ACI_Midrange ~ Tot_Knocks + Num_Herbivory + Num_L_calls + Site +
## Hour + Year + Tot_Knocks:Site + Tot_Knocks:Hour + Num_Herbivory:Site +
## Num_Herbivory:Hour + Num_L_calls:Hour
##
## Df Deviance AIC
## - Tot_Knocks:Hour 3 3.8849 4042.9
## - Tot_Knocks:Site 4 3.9547 4043.7
## - Num_Herbivory:Hour 3 3.9652 4046.1
## - Num_L_calls:Hour 3 3.9659 4046.1
## - Num_Herbivory:Site 4 4.0246 4046.5
## - Year 1 3.8747 4046.5
## <none> 3.8643 4048.1
##
## Step: AIC=4043.1
## ACI_Midrange ~ Tot_Knocks + Num_Herbivory + Num_L_calls + Site +
## Hour + Year + Tot_Knocks:Site + Num_Herbivory:Site + Num_Herbivory:Hour +
## Num_L_calls:Hour
##
## Df Deviance AIC
## - Tot_Knocks:Site 4 3.9597 4038.1
## - Num_Herbivory:Hour 3 3.9721 4040.6
## - Num_L_calls:Hour 3 3.9864 4041.2
## - Num_Herbivory:Site 4 4.0413 4041.4
## - Year 1 3.8959 4041.5
## <none> 3.8849 4043.1
##
## Step: AIC=4038.76
## ACI_Midrange ~ Tot_Knocks + Num_Herbivory + Num_L_calls + Site +
## Hour + Year + Num_Herbivory:Site + Num_Herbivory:Hour + Num_L_calls:Hour
##
## Df Deviance AIC
## - Num_Herbivory:Hour 3 4.0359 4035.9
## - Num_L_calls:Hour 3 4.0467 4036.3
## - Year 1 3.9627 4036.9
## - Num_Herbivory:Site 4 4.1360 4037.9
## - Tot_Knocks 1 3.9995 4038.4
## <none> 3.9597 4038.8
##
## Step: AIC=4036.41
## ACI_Midrange ~ Tot_Knocks + Num_Herbivory + Num_L_calls + Site +
## Hour + Year + Num_Herbivory:Site + Num_L_calls:Hour
##
## Df Deviance AIC
## - Num_L_calls:Hour 3 4.1075 4033.3
## - Year 1 4.0389 4034.5
## - Num_Herbivory:Site 4 4.2302 4036.3
## <none> 4.0359 4036.4
## - Tot_Knocks 1 4.0871 4036.5
##
## Step: AIC=4033.78
## ACI_Midrange ~ Tot_Knocks + Num_Herbivory + Num_L_calls + Site +
## Hour + Year + Num_Herbivory:Site
##
## Df Deviance AIC
## - Hour 3 4.1672 4030.2
## - Year 1 4.1094 4031.9
## - Num_Herbivory:Site 4 4.3010 4033.6
## <none> 4.1075 4033.8
## - Tot_Knocks 1 4.1576 4033.8
## - Num_L_calls 1 4.1776 4034.6
##
## Step: AIC=4030.55
## ACI_Midrange ~ Tot_Knocks + Num_Herbivory + Num_L_calls + Site +
## Year + Num_Herbivory:Site
##
## Df Deviance AIC
## - Year 1 4.1686 4028.6
## - Tot_Knocks 1 4.2086 4030.2
## - Num_L_calls 1 4.2113 4030.3
## <none> 4.1672 4030.5
## - Num_Herbivory:Site 4 4.3652 4030.6
##
## Step: AIC=4028.61
## ACI_Midrange ~ Tot_Knocks + Num_Herbivory + Num_L_calls + Site +
## Num_Herbivory:Site
##
## Df Deviance AIC
## - Tot_Knocks 1 4.2092 4028.3
## - Num_L_calls 1 4.2137 4028.4
## <none> 4.1686 4028.6
## - Num_Herbivory:Site 4 4.3721 4028.9
##
## Step: AIC=4028.47
## ACI_Midrange ~ Num_Herbivory + Num_L_calls + Site + Num_Herbivory:Site
##
## Df Deviance AIC
## - Num_L_calls 1 4.2572 4028.4
## <none> 4.2092 4028.5
## - Num_Herbivory:Site 4 4.4228 4029.2
##
## Step: AIC=4028.64
## ACI_Midrange ~ Num_Herbivory + Site + Num_Herbivory:Site
##
## Call: glm(formula = ACI_Midrange ~ Num_Herbivory + Site + Num_Herbivory:Site,
## family = "Gamma", data = AC.DF1Co)
##
## Coefficients:
## (Intercept) Num_Herbivory Site35
## 1.795e-05 3.282e-07 -1.544e-06
## Site40 Site5 Site8
## 9.005e-07 -1.530e-06 -1.156e-06
## Num_Herbivory:Site35 Num_Herbivory:Site40 Num_Herbivory:Site5
## -2.453e-07 -3.617e-07 -3.830e-07
## Num_Herbivory:Site8
## -3.881e-07
##
## Degrees of Freedom: 190 Total (i.e. Null); 181 Residual
## Null Deviance: 4.987
## Residual Deviance: 4.257 AIC: 4029
AICc(fit.a, ACI.mf.lm)
## df AICc
## fit.a 34 4070.423
## ACI.mf.lm 4 4038.392
Most parsimonious and best model from the AIC stepwise model selection
#ACI_Midrange ~ Tot_Knocks + Num_Herbivory + Num_L_calls + Site + Hour + Year + Num_Herbivory:Site
fit.a2 <- glm(ACI_Midrange ~ Tot_Knocks + Num_Herbivory + Num_L_calls + Site + Hour + Num_Herbivory:Site, data = AC.DF1Co, family = "Gamma")
summary(fit.a2)
##
## Call:
## glm(formula = ACI_Midrange ~ Tot_Knocks + Num_Herbivory + Num_L_calls +
## Site + Hour + Num_Herbivory:Site, family = "Gamma", data = AC.DF1Co)
##
## Deviance Residuals:
## Min 1Q Median 3Q Max
## -0.24565 -0.13928 -0.02445 0.09444 0.38612
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 1.765e-05 1.247e-06 14.159 <2e-16 ***
## Tot_Knocks -5.791e-09 4.065e-09 -1.424 0.1561
## Num_Herbivory 3.515e-07 4.013e-07 0.876 0.3823
## Num_L_calls -5.311e-08 3.028e-08 -1.754 0.0811 .
## Site35 -1.617e-06 1.255e-06 -1.288 0.1993
## Site40 1.023e-06 1.277e-06 0.801 0.4244
## Site5 -1.279e-06 1.285e-06 -0.995 0.3209
## Site8 -1.117e-06 1.241e-06 -0.900 0.3692
## Hour21 6.821e-07 6.162e-07 1.107 0.2699
## Hour3 4.663e-07 6.153e-07 0.758 0.4495
## Hour9 -8.317e-08 6.137e-07 -0.136 0.8924
## Num_Herbivory:Site35 -2.689e-07 4.043e-07 -0.665 0.5069
## Num_Herbivory:Site40 -3.351e-07 4.185e-07 -0.801 0.4244
## Num_Herbivory:Site5 -4.007e-07 4.039e-07 -0.992 0.3225
## Num_Herbivory:Site8 -4.079e-07 4.026e-07 -1.013 0.3123
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for Gamma family taken to be 0.02451643)
##
## Null deviance: 4.9867 on 190 degrees of freedom
## Residual deviance: 4.1094 on 176 degrees of freedom
## AIC: 4031.9
##
## Number of Fisher Scoring iterations: 4
ggplot(data =Snap.HF, aes(Snap.HF$ACI_HF)) + geom_histogram() + ggtitle("HF ACI distribution") + geom_vline(aes(xintercept = mean(ACI_HF)), color = "red", linetype ="dashed", size = 1) + geom_vline(aes(xintercept = median(ACI_HF)), color = "blue", linetype = "dotted", size = 1)
## `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.
Distribution looks normal, given what we discussed about the HF SPL distribution
#testing time splits for this model to confirm they are the same as SPL HF model
afit.tg <- lm(ACI_HF ~ Snaps*tg*Site + Year, data = Snap.HFC)
afit.dn <- lm(ACI_HF ~ Snaps*dn*Site + Year, data = Snap.HFC)
afit.ns <- lm(ACI_HF ~ Snaps*ns*Site + Year, data = Snap.HFC)
afit.t12 <- lm(ACI_HF ~Snaps*t12*Site + Year, data = Snap.HFC)
AICc(afit.tg, afit.dn, afit.ns, afit.t12)
## df AICc
## afit.tg 42 318203.3
## afit.dn 22 318276.9
## afit.ns 32 318292.4
## afit.t12 22 318159.7
fit.hfaci <- lm(ACI_HF ~ Snaps*t12 + Snaps*Site + t12*Site + Year, data = Snap.HFC)
fit.hfaci2 <- lm(ACI_HF ~ Snaps*t12*Site + Year, data = Snap.HFC)
fit.hfaci3 <- lm(ACI_HF ~ Snaps*t12 + Snaps*Site + Year, data = Snap.HFC)
AICc(fit.hfaci,fit.hfaci2, fit.hfaci3)
## df AICc
## fit.hfaci 18 318164.8
## fit.hfaci2 22 318159.7
## fit.hfaci3 14 318393.5
summary(fit.hfaci2)
##
## Call:
## lm(formula = ACI_HF ~ Snaps * t12 * Site + Year, data = Snap.HFC)
##
## Residuals:
## Min 1Q Median 3Q Max
## -12613.4 -3641.7 -958.7 2729.3 21530.1
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) -2.883e+05 1.821e+05 -1.583 0.11351
## Snaps -5.588e+00 2.739e+00 -2.040 0.04135 *
## t12Low -1.203e+03 1.866e+02 -6.445 1.19e-10 ***
## Site35 4.440e+03 2.333e+02 19.033 < 2e-16 ***
## Site40 3.104e+01 2.469e+02 0.126 0.89996
## Site5 4.635e+02 1.972e+02 2.350 0.01880 *
## Site8 3.123e+03 1.827e+02 17.096 < 2e-16 ***
## Year 1.739e+02 9.029e+01 1.927 0.05406 .
## Snaps:t12Low 8.336e+00 3.251e+00 2.564 0.01034 *
## Snaps:Site35 9.194e-01 4.078e+00 0.225 0.82164
## Snaps:Site40 -7.411e+00 4.369e+00 -1.696 0.08987 .
## Snaps:Site5 1.576e+00 4.318e+00 0.365 0.71516
## Snaps:Site8 -1.934e+01 3.950e+00 -4.896 9.85e-07 ***
## t12Low:Site35 5.579e+02 3.166e+02 1.762 0.07806 .
## t12Low:Site40 2.276e+02 3.241e+02 0.702 0.48259
## t12Low:Site5 3.701e+03 2.738e+02 13.519 < 2e-16 ***
## t12Low:Site8 -4.602e+02 3.445e+02 -1.336 0.18169
## Snaps:t12Low:Site35 -6.317e+00 5.227e+00 -1.209 0.22683
## Snaps:t12Low:Site40 -9.996e+00 5.456e+00 -1.832 0.06693 .
## Snaps:t12Low:Site5 -1.329e+01 5.584e+00 -2.380 0.01733 *
## Snaps:t12Low:Site8 -1.707e+01 5.080e+00 -3.360 0.00078 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 5069 on 15966 degrees of freedom
## Multiple R-squared: 0.1587, Adjusted R-squared: 0.1577
## F-statistic: 150.6 on 20 and 15966 DF, p-value: < 2.2e-16
stepAIC(fit.hfaci2)
## Start: AIC=272788.5
## ACI_HF ~ Snaps * t12 * Site + Year
##
## Df Sum of Sq RSS AIC
## <none> 4.1026e+11 272788
## - Year 1 95367893 4.1035e+11 272790
## - Snaps:t12:Site 4 338278009 4.1059e+11 272794
##
## Call:
## lm(formula = ACI_HF ~ Snaps * t12 * Site + Year, data = Snap.HFC)
##
## Coefficients:
## (Intercept) Snaps t12Low
## -2.883e+05 -5.587e+00 -1.203e+03
## Site35 Site40 Site5
## 4.440e+03 3.104e+01 4.635e+02
## Site8 Year Snaps:t12Low
## 3.123e+03 1.739e+02 8.336e+00
## Snaps:Site35 Snaps:Site40 Snaps:Site5
## 9.194e-01 -7.411e+00 1.576e+00
## Snaps:Site8 t12Low:Site35 t12Low:Site40
## -1.934e+01 5.579e+02 2.276e+02
## t12Low:Site5 t12Low:Site8 Snaps:t12Low:Site35
## 3.701e+03 -4.602e+02 -6.317e+00
## Snaps:t12Low:Site40 Snaps:t12Low:Site5 Snaps:t12Low:Site8
## -9.996e+00 -1.329e+01 -1.707e+01
IP4 <- interaction.plot(Snap.HFC$t12, Snap.HFC$Site, Snap.HF$ACI_HF)
So it looks like ACI has a significant 3 way interaction at site 5… WHAT DOES THIS MEAN AND HOW DO I SHOW IT
Show it - I think I can make a 2 frame plot, facet_wrap by time, showing the effect between Snaps and ACI in each plot
What does it mean - it means that HF ACI is significantly associated with combined changes of Snaps and Time and Site (but only at site 5?)
#fit.hfaci <- lm(ACI_HF ~(Snaps*t12*Site) + Year, data = Snap.HF)
#summary(fit.hfaci)
#stepAIC(fit.a)
#AICc(fit.a, ACI.mf.lm)
Determining which is the best way to group the snaps by time
tg = quarters (00-05, 06-11, 12-17, 18-23) dn = day night (18-05, 6-17) ns = nine cycle (22-03, 04-09, 10-15, 16-21) t12 = my half and half cycle (2140 - 920, 920 - 2140)
fit.tg <- lm(SPL_HF ~ Snaps*tg*Site + Year, data = Snap.HFC)
fit.dn <- lm(SPL_HF ~ Snaps*dn*Site + Year, data = Snap.HFC)
fit.ns <- lm(SPL_HF ~ Snaps*ns*Site + Year, data = Snap.HFC)
fit.t12 <- lm(SPL_HF ~Snaps*t12*Site + Year, data = Snap.HFC)
#models that have a 2 way interaction and time seperately as a factor
fit.t12t <- lm(SPL_HF ~ Snaps*Site + t12 + Year, data = Snap.HFC)
fit.tgt <- lm(SPL_HF ~ Snaps*Site + tg + Year, data = Snap.HFC)
fit.dnt <- lm(SPL_HF ~ Snaps*Site + dn + Year, data = Snap.HFC)
fit.nst <- lm(SPL_HF ~ Snaps*Site + ns + Year, data = Snap.HFC)
AICc(fit.tg, fit.dn, fit.ns, fit.t12, fit.t12t, fit.tgt, fit.dnt, fit.nst)
## df AICc
## fit.tg 42 68564.26
## fit.dn 22 72450.05
## fit.ns 32 70408.03
## fit.t12 22 64311.29
## fit.t12t 13 65331.35
## fit.tgt 15 70314.51
## fit.dnt 13 72650.42
## fit.nst 14 71856.06
Best model was the one that split time at 9:20 and 21:40
Confused about my next steps here
stepAIC(fit.t12)
## Start: AIC=18940.08
## SPL_HF ~ Snaps * t12 * Site + Year
##
## Df Sum of Sq RSS AIC
## <none> 52137 18940
## - Snaps:t12:Site 4 1087 53223 19262
## - Year 1 35884 88021 27311
##
## Call:
## lm(formula = SPL_HF ~ Snaps * t12 * Site + Year, data = Snap.HFC)
##
## Coefficients:
## (Intercept) Snaps t12Low
## -6.687e+03 -8.989e-03 -2.886e+00
## Site35 Site40 Site5
## -1.421e+00 -1.297e+00 -1.782e+00
## Site8 Year Snaps:t12Low
## 1.615e+00 3.374e+00 5.547e-03
## Snaps:Site35 Snaps:Site40 Snaps:Site5
## 3.703e-02 1.997e-02 8.580e-03
## Snaps:Site8 t12Low:Site35 t12Low:Site40
## 5.426e-03 -7.920e-02 -8.341e-01
## t12Low:Site5 t12Low:Site8 Snaps:t12Low:Site35
## -1.815e+00 -2.476e+00 -3.223e-02
## Snaps:t12Low:Site40 Snaps:t12Low:Site5 Snaps:t12Low:Site8
## -7.714e-03 -2.101e-03 -9.700e-03
#returned only the three way interaction - so I am going to try some manual selection to see if there is a more parsimonious model
fit.hf1 <- lm(SPL_HF ~ Snaps + t12 + Site + Snaps:t12 + Snaps:Site + t12:Site, data = Snap.HFC)
fit.hf2 <- lm(SPL_HF ~ Snaps + t12 + Site + Snaps:t12 + Snaps:Site, data = Snap.HFC)
fit.hf3 <- lm(SPL_HF ~ Snaps + Site + Snaps:Site, data = Snap.HFC)
AICc(fit.t12,fit.hf1, fit.hf2, fit.hf3)
## df AICc
## fit.t12 22 64311.29
## fit.hf1 17 73122.38
## fit.hf2 13 73615.19
## fit.hf3 11 78480.77
#SPL_HF ~ Snaps + t12 + Site + Snaps:t12 + Snaps:Site + t12:Site
snap.model <- lm(SPL_HF ~ Snaps + t12 + Site + Snaps:t12 + Snaps:Site + t12:Site, data = Snap.HFC)
Anova(fit.t12, type = 3)
## Anova Table (Type III tests)
##
## Response: SPL_HF
## Sum Sq Df F value Pr(>F)
## (Intercept) 34636 1 10606.698 < 2.2e-16 ***
## Snaps 277 1 84.770 < 2.2e-16 ***
## t12 6147 1 1882.447 < 2.2e-16 ***
## Site 9383 4 718.363 < 2.2e-16 ***
## Year 35884 1 10989.000 < 2.2e-16 ***
## Snaps:t12 75 1 22.915 1.709e-06 ***
## Snaps:Site 2628 4 201.172 < 2.2e-16 ***
## t12:Site 2166 4 165.863 < 2.2e-16 ***
## Snaps:t12:Site 1087 4 83.194 < 2.2e-16 ***
## Residuals 52137 15966
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
summary(fit.t12)
##
## Call:
## lm(formula = SPL_HF ~ Snaps * t12 * Site + Year, data = Snap.HFC)
##
## Residuals:
## Min 1Q Median 3Q Max
## -6.0059 -1.1523 -0.0511 1.0325 9.5333
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) -6.687e+03 6.493e+01 -102.989 < 2e-16 ***
## Snaps -8.989e-03 9.763e-04 -9.207 < 2e-16 ***
## t12Low -2.886e+00 6.652e-02 -43.387 < 2e-16 ***
## Site35 -1.421e+00 8.317e-02 -17.087 < 2e-16 ***
## Site40 -1.297e+00 8.802e-02 -14.738 < 2e-16 ***
## Site5 -1.782e+00 7.031e-02 -25.344 < 2e-16 ***
## Site8 1.615e+00 6.512e-02 24.808 < 2e-16 ***
## Year 3.374e+00 3.219e-02 104.828 < 2e-16 ***
## Snaps:t12Low 5.547e-03 1.159e-03 4.787 1.71e-06 ***
## Snaps:Site35 3.703e-02 1.454e-03 25.471 < 2e-16 ***
## Snaps:Site40 1.997e-02 1.558e-03 12.819 < 2e-16 ***
## Snaps:Site5 8.580e-03 1.539e-03 5.574 2.52e-08 ***
## Snaps:Site8 5.426e-03 1.408e-03 3.853 0.000117 ***
## t12Low:Site35 -7.920e-02 1.129e-01 -0.702 0.482821
## t12Low:Site40 -8.341e-01 1.155e-01 -7.219 5.49e-13 ***
## t12Low:Site5 -1.815e+00 9.759e-02 -18.596 < 2e-16 ***
## t12Low:Site8 -2.476e+00 1.228e-01 -20.156 < 2e-16 ***
## Snaps:t12Low:Site35 -3.223e-02 1.863e-03 -17.297 < 2e-16 ***
## Snaps:t12Low:Site40 -7.714e-03 1.945e-03 -3.967 7.32e-05 ***
## Snaps:t12Low:Site5 -2.101e-03 1.991e-03 -1.056 0.291199
## Snaps:t12Low:Site8 -9.700e-03 1.811e-03 -5.357 8.59e-08 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 1.807 on 15966 degrees of freedom
## Multiple R-squared: 0.7323, Adjusted R-squared: 0.7319
## F-statistic: 2183 on 20 and 15966 DF, p-value: < 2.2e-16
plot(fit.t12)
hist(resid(snap.model))
## NULL
## NULL
## NULL
#SPL_Midrange ~ Tot_Knocks + Num_Herbivory + Num_L_calls + Site + Hour + Year + Tot_Knocks:Site + Tot_Knocks:Hour + Num_Herbivory:Site + Num_L_calls:Site + Num_L_calls:Hour
fit.r <- lmer(SPL_Midrange ~ Tot_Knocks + Num_Herbivory + Num_L_calls + (1|Site) + Hour + Year + Tot_Knocks:Hour + Num_Herbivory:Hour + Num_L_calls:Hour, data = AC.DF1Co)
Anova(fit.r, type=3)
## Analysis of Deviance Table (Type III Wald chisquare tests)
##
## Response: SPL_Midrange
## Chisq Df Pr(>Chisq)
## (Intercept) 26167.2057 1 < 2.2e-16 ***
## Tot_Knocks 4.4084 1 0.035762 *
## Num_Herbivory 15.1849 1 9.748e-05 ***
## Num_L_calls 0.0549 1 0.814734
## Hour 8.5280 3 0.036271 *
## Year 134.8468 1 < 2.2e-16 ***
## Tot_Knocks:Hour 39.2710 3 1.521e-08 ***
## Num_Herbivory:Hour 5.6442 3 0.130262
## Num_L_calls:Hour 11.4313 3 0.009608 **
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
summary(fit.r)
## Linear mixed model fit by REML. t-tests use Satterthwaite's method [
## lmerModLmerTest]
## Formula: SPL_Midrange ~ Tot_Knocks + Num_Herbivory + Num_L_calls + (1 |
## Site) + Hour + Year + Tot_Knocks:Hour + Num_Herbivory:Hour +
## Num_L_calls:Hour
## Data: AC.DF1Co
##
## REML criterion at convergence: 881.4
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -1.96637 -0.64984 -0.04602 0.52505 2.61169
##
## Random effects:
## Groups Name Variance Std.Dev.
## Site (Intercept) 0.9442 0.9717
## Residual 4.5753 2.1390
## Number of obs: 191, groups: Site, 5
##
## Fixed effects:
## Estimate Std. Error df t value Pr(>|t|)
## (Intercept) 102.488770 0.633574 12.859440 161.763 < 2e-16 ***
## Tot_Knocks -0.019041 0.009069 173.992210 -2.100 0.037204 *
## Num_Herbivory 0.089278 0.022911 172.081483 3.897 0.000139 ***
## Num_L_calls 0.021623 0.092279 172.051789 0.234 0.815013
## Hour21 1.316661 0.545335 171.430449 2.414 0.016813 *
## Hour3 -1.934212 2.147264 171.274496 -0.901 0.368970
## Hour9 2.664335 2.029130 170.480181 1.313 0.190933
## Year18 3.654715 0.314726 170.150269 11.612 < 2e-16 ***
## Tot_Knocks:Hour21 0.023039 0.010394 170.954532 2.217 0.027969 *
## Tot_Knocks:Hour3 0.023985 0.010142 172.549443 2.365 0.019149 *
## Tot_Knocks:Hour9 0.058132 0.010404 172.825185 5.587 8.83e-08 ***
## Num_Herbivory:Hour21 -0.137976 0.070233 170.741418 -1.965 0.051089 .
## Num_Herbivory:Hour3 -0.894827 0.707154 170.897624 -1.265 0.207453
## Num_Herbivory:Hour9 0.274713 0.687808 170.202108 0.399 0.690097
## Num_L_calls:Hour21 0.075658 0.095433 171.058180 0.793 0.428998
## Num_L_calls:Hour3 0.018287 0.118594 170.022168 0.154 0.877637
## Num_L_calls:Hour9 -0.274625 0.137256 172.933191 -2.001 0.046976 *
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation matrix not shown by default, as p = 17 > 12.
## Use print(x, correlation=TRUE) or
## vcov(x) if you need it
plot(fit.r)
This model will not work with site as a random effect right now, trying to figure out the source of it still.
#ACI_Midrange ~ Tot_Knocks + Num_Herbivory + Num_L_calls + Site + Hour + Year + Num_Herbivory:Site
fit.a2 <- glm(ACI_Midrange ~ Tot_Knocks + Num_Herbivory + Num_L_calls + Site + Hour + Num_Herbivory:Site , data = AC.DF1Co, family = "Gamma")
summary(fit.a2)
##
## Call:
## glm(formula = ACI_Midrange ~ Tot_Knocks + Num_Herbivory + Num_L_calls +
## Site + Hour + Num_Herbivory:Site, family = "Gamma", data = AC.DF1Co)
##
## Deviance Residuals:
## Min 1Q Median 3Q Max
## -0.24565 -0.13928 -0.02445 0.09444 0.38612
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 1.765e-05 1.247e-06 14.159 <2e-16 ***
## Tot_Knocks -5.791e-09 4.065e-09 -1.424 0.1561
## Num_Herbivory 3.515e-07 4.013e-07 0.876 0.3823
## Num_L_calls -5.311e-08 3.028e-08 -1.754 0.0811 .
## Site35 -1.617e-06 1.255e-06 -1.288 0.1993
## Site40 1.023e-06 1.277e-06 0.801 0.4244
## Site5 -1.279e-06 1.285e-06 -0.995 0.3209
## Site8 -1.117e-06 1.241e-06 -0.900 0.3692
## Hour21 6.821e-07 6.162e-07 1.107 0.2699
## Hour3 4.663e-07 6.153e-07 0.758 0.4495
## Hour9 -8.317e-08 6.137e-07 -0.136 0.8924
## Num_Herbivory:Site35 -2.689e-07 4.043e-07 -0.665 0.5069
## Num_Herbivory:Site40 -3.351e-07 4.185e-07 -0.801 0.4244
## Num_Herbivory:Site5 -4.007e-07 4.039e-07 -0.992 0.3225
## Num_Herbivory:Site8 -4.079e-07 4.026e-07 -1.013 0.3123
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for Gamma family taken to be 0.02451643)
##
## Null deviance: 4.9867 on 190 degrees of freedom
## Residual deviance: 4.1094 on 176 degrees of freedom
## AIC: 4031.9
##
## Number of Fisher Scoring iterations: 4
Distribution looks normal, given what we discussed about the HF SPL distribution
#testing time splits for this model to confirm they are the same as SPL HF model
afit.tg1 <- lmer(ACI_HF ~ Snaps*tg + (1|Site) + Year, data = Snap.HFC)
afit.dn1 <- lmer(ACI_HF ~ Snaps*dn + (1|Site) + Year, data = Snap.HFC)
afit.ns1 <- lmer(ACI_HF ~ Snaps*ns + (1|Site) + Year, data = Snap.HFC)
afit.t121 <- lmer(ACI_HF ~ Snaps*t12 + (1|Site) + Year, data = Snap.HFC)
AICc(afit.tg1, afit.dn1, afit.ns1, afit.t121)
## df AICc
## afit.tg1 11 318500.2
## afit.dn1 7 318534.5
## afit.ns1 9 318517.5
## afit.t121 7 318535.6
summary(afit.ns1)
## Linear mixed model fit by REML. t-tests use Satterthwaite's method [
## lmerModLmerTest]
## Formula: ACI_HF ~ Snaps * ns + (1 | Site) + Year
## Data: Snap.HFC
##
## REML criterion at convergence: 318499.5
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -2.4291 -0.7243 -0.1895 0.5481 4.4411
##
## Random effects:
## Groups Name Variance Std.Dev.
## Site (Intercept) 4330194 2081
## Residual 26336484 5132
## Number of obs: 15987, groups: Site, 5
##
## Fixed effects:
## Estimate Std. Error df t value Pr(>|t|)
## (Intercept) 2.422e+05 1.705e+05 1.598e+04 1.421 0.155
## Snaps -9.079e+00 1.743e+00 1.598e+04 -5.210 1.91e-07 ***
## ns9AM-3PM -1.664e+02 1.240e+02 1.598e+04 -1.342 0.180
## ns9PM-3AM 1.055e+02 1.081e+02 1.598e+04 0.976 0.329
## Year -8.819e+01 8.449e+01 1.598e+04 -1.044 0.297
## Snaps:ns9AM-3PM 2.043e-01 2.189e+00 1.598e+04 0.093 0.926
## Snaps:ns9PM-3AM -2.970e-01 1.952e+00 1.598e+04 -0.152 0.879
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation of Fixed Effects:
## (Intr) Snaps n9AM-3 n9PM-3 Year S:9AM-
## Snaps 0.021
## ns9AM-3PM 0.011 0.343
## ns9PM-3AM 0.010 0.392 0.624
## Year -1.000 -0.021 -0.012 -0.010
## Snp:9AM-3PM -0.040 -0.788 -0.164 -0.308 0.040
## Snp:9PM-3AM -0.017 -0.877 -0.297 -0.326 0.017 0.704
Determining which is the best way to group the snaps by time
tg = quarters (00-05, 06-11, 12-17, 18-23) dn = day night (18-05, 6-17) ns = nine cycle (22-03, 04-09, 10-15, 16-21) t12 = my half and half cycle (2140 - 920, 920 - 2140)
fit.rtg <- lmer(SPL_HF ~ Snaps*tg + (1|Site) + Year, data = Snap.HFC)
fit.rdn <- lmer(SPL_HF ~ Snaps*dn + (1|Site) + Year, data = Snap.HFC)
fit.rns <- lmer(SPL_HF ~ Snaps*ns + (1|Site) + Year, data = Snap.HFC)
fit.rt12 <- lmer(SPL_HF ~Snaps*t12 + (1|Site) + Year, data = Snap.HFC)
#models that have a 2 way interaction and time seperately as a factor
AICc(fit.rtg, fit.rdn, fit.rns, fit.rt12)
## df AICc
## fit.rtg 11 71334.97
## fit.rdn 7 74361.55
## fit.rns 9 72857.07
## fit.rt12 7 65738.30
Best model was the one that split time at 9:20 and 21:40
Confused about my next steps here
#returned only the three way interaction - so I am going to try some manual selection to see if there is a more parsimonious model
#SPL_HF ~ Snaps + t12 + Site + Snaps:t12 + Snaps:Site + t12:Site
Anova(fit.rt12, type = 3)
## Analysis of Deviance Table (Type III Wald chisquare tests)
##
## Response: SPL_HF
## Chisq Df Pr(>Chisq)
## (Intercept) 11516.65 1 < 2.2e-16 ***
## Snaps 120.20 1 < 2.2e-16 ***
## t12 14944.85 1 < 2.2e-16 ***
## Year 11926.88 1 < 2.2e-16 ***
## Snaps:t12 16.98 1 3.778e-05 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
summary(fit.rt12)
## Linear mixed model fit by REML. t-tests use Satterthwaite's method [
## lmerModLmerTest]
## Formula: SPL_HF ~ Snaps * t12 + (1 | Site) + Year
## Data: Snap.HFC
##
## REML criterion at convergence: 65724.3
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -3.1864 -0.6744 -0.0032 0.5999 5.0895
##
## Random effects:
## Groups Name Variance Std.Dev.
## Site (Intercept) 1.884 1.373
## Residual 3.557 1.886
## Number of obs: 15987, groups: Site, 5
##
## Fixed effects:
## Estimate Std. Error df t value Pr(>|t|)
## (Intercept) -6.760e+03 6.299e+01 1.598e+04 -107.316 < 2e-16 ***
## Snaps 5.145e-03 4.693e-04 1.598e+04 10.964 < 2e-16 ***
## t12Low -4.076e+00 3.334e-02 1.598e+04 -122.249 < 2e-16 ***
## Year 3.410e+00 3.122e-02 1.598e+04 109.210 < 2e-16 ***
## Snaps:t12Low -2.492e-03 6.048e-04 1.598e+04 -4.121 3.8e-05 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation of Fixed Effects:
## (Intr) Snaps t12Low Year
## Snaps 0.087
## t12Low 0.021 0.353
## Year -1.000 -0.087 -0.022
## Snaps:t12Lw -0.113 -0.793 -0.114 0.113
plot(fit.rt12)
Here I am going to build my models one step at a time, interpretting their results each time before I take the next step and make them more complex.
##
## Call:
## lm(formula = SPL_HF ~ Snaps + Year, data = Snap.HF)
##
## Residuals:
## Min 1Q Median 3Q Max
## -8.7519 -2.0102 0.0929 2.0136 13.5007
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 9.582e+01 5.598e-01 171.17 <2e-16 ***
## Snaps 1.372e-02 3.828e-04 35.84 <2e-16 ***
## Year2018 3.418e+00 4.881e-02 70.02 <2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 2.969 on 15984 degrees of freedom
## Multiple R-squared: 0.2762, Adjusted R-squared: 0.2761
## F-statistic: 3050 on 2 and 15984 DF, p-value: < 2.2e-16
Right now what this means is that for each snap produced HF SPL increases by .0137 and that in 2018 the SPL was on average 3.418 dB louder than in 2017. However, this only explains 27% of the variation in HF SPL.
##
## Call:
## lm(formula = SPL_HF ~ Snaps + t12 + Year, data = Snap.HF)
##
## Residuals:
## Min 1Q Median 3Q Max
## -6.7528 -1.5323 -0.0308 1.4807 11.1021
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 1.189e+02 4.693e-01 253.44 <2e-16 ***
## Snaps -6.209e-04 3.153e-04 -1.97 0.0489 *
## t12Low -4.297e+00 3.869e-02 -111.05 <2e-16 ***
## Year2018 3.441e+00 3.668e-02 93.82 <2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 2.231 on 15983 degrees of freedom
## Multiple R-squared: 0.5915, Adjusted R-squared: 0.5914
## F-statistic: 7713 on 3 and 15983 DF, p-value: < 2.2e-16
Here we can see that the relationship between snaps and SPL was actually a part of the day/night cycle in SPL. Therefore, adding that to the model accounts for that variability and instead we get that snaps actually have a very small effect size and a negative one at that, (-0.00062).
We do pull apart the effect of time of day though, with daytime(t12low) having an SPL that is 4.297 dB lower than night time SPL.
The addition of time didn’t pull largely away from the effect of year as well, as 2018 still seemed to have an SPL that was 3.441 dB higher than 2017.
## Linear mixed model fit by REML. t-tests use Satterthwaite's method [
## lmerModLmerTest]
## Formula: SPL_HF ~ Snaps + t12 + Year + (1 | Site)
## Data: Snap.HF
##
## REML criterion at convergence: 65728.3
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -3.1941 -0.6737 -0.0087 0.5984 5.0594
##
## Random effects:
## Groups Name Variance Std.Dev.
## Site (Intercept) 1.870 1.367
## Residual 3.561 1.887
## Number of obs: 15987, groups: Site, 5
##
## Fixed effects:
## Estimate Std. Error df t value Pr(>|t|)
## (Intercept) 1.127e+02 7.450e-01 8.798e+00 151.23 2.48e-16 ***
## Snaps 3.612e-03 2.862e-04 1.598e+04 12.62 < 2e-16 ***
## t12Low -4.092e+00 3.314e-02 1.598e+04 -123.45 < 2e-16 ***
## Year2018 3.425e+00 3.104e-02 1.598e+04 110.33 < 2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation of Fixed Effects:
## (Intr) Snaps t12Low
## Snaps -0.570
## t12Low -0.266 0.434
## Year2018 -0.018 0.005 -0.009
When we add site to our model as a random effect, it takes care of variance that was obscuring the effect of snaps on SPL. Therefore, with this we get an increase in the effect size of snaps on SPL, coming back to a logically positive number that is still quite small (0.003612).
We also get a slightly different effect, but more or less the same direction of an effect by t12 (day is now only 4 dB lower, meaning some of the variation is taken up by site). Some of the variation between 2017 and 2018 was also taken by site as the effect size has also shrunk a tiny bit from 3.441 to 3.425.
## Linear mixed model fit by REML. t-tests use Satterthwaite's method [
## lmerModLmerTest]
## Formula: SPL_HF ~ Snaps * t12 + Year + (1 | Site)
## Data: Snap.HF
##
## REML criterion at convergence: 65724.3
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -3.1864 -0.6744 -0.0032 0.5999 5.0895
##
## Random effects:
## Groups Name Variance Std.Dev.
## Site (Intercept) 1.884 1.373
## Residual 3.557 1.886
## Number of obs: 15987, groups: Site, 5
##
## Fixed effects:
## Estimate Std. Error df t value Pr(>|t|)
## (Intercept) 1.104e+02 9.283e-01 2.089e+01 118.923 < 2e-16 ***
## Snaps 5.145e-03 4.693e-04 1.598e+04 10.964 < 2e-16 ***
## t12Low -4.383e-01 8.872e-01 1.598e+04 -0.494 0.621
## Year2018 3.410e+00 3.122e-02 1.598e+04 109.210 < 2e-16 ***
## Snaps:t12Low -2.492e-03 6.048e-04 1.598e+04 -4.121 3.8e-05 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation of Fixed Effects:
## (Intr) Snaps t12Low Yr2018
## Snaps -0.750
## t12Low -0.602 0.802
## Year2018 0.053 -0.087 -0.114
## Snaps:t12Lw 0.594 -0.793 -0.999 0.113
Now when we add a two way interaction between Snaps and time of day, we are looking to see if time of day changes the relationship between snaps and HF SPL. It is significant, and what it says is that during the day the relationship between snaps and SPL is reduced from 0.005145 to 0.005145 - 0.002492 = 0.002653. This means that during the day, the effect of snaps on HF SPL is smaller than it is at night.
##
## Call:
## lm(formula = SPL_Midrange ~ Tot_Knocks, data = AC.DF1Co)
##
## Residuals:
## Min 1Q Median 3Q Max
## -7.2088 -2.1924 -0.7869 1.6962 11.9299
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 1.058e+02 2.389e-01 442.680 < 2e-16 ***
## Tot_Knocks 1.801e-02 4.252e-03 4.237 3.54e-05 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 3.302 on 189 degrees of freedom
## Multiple R-squared: 0.08674, Adjusted R-squared: 0.08191
## F-statistic: 17.95 on 1 and 189 DF, p-value: 3.538e-05
Right now this means that each knock increases SPL by 0.018.
##
## Call:
## lm(formula = SPL_Midrange ~ Tot_Knocks + Num_Herbivory + Num_L_calls,
## data = AC.DF1Co)
##
## Residuals:
## Min 1Q Median 3Q Max
## -7.0510 -2.1997 -0.7752 1.7692 12.0918
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 1.058e+02 2.383e-01 443.875 < 2e-16 ***
## Tot_Knocks 1.796e-02 4.258e-03 4.217 3.85e-05 ***
## Num_Herbivory 4.951e-02 2.862e-02 1.730 0.0853 .
## Num_L_calls 1.577e-03 3.457e-02 0.046 0.9637
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 3.293 on 187 degrees of freedom
## Multiple R-squared: 0.1013, Adjusted R-squared: 0.08685
## F-statistic: 7.024 on 3 and 187 DF, p-value: 0.0001682
The addition of Herbivory and Long calls did little as they were insignificant, however, you can see their addition had a tiny effect by shrinking the effect size of knocks to 0.0179.
##
## Call:
## lm(formula = SPL_Midrange ~ Tot_Knocks + Num_Herbivory + Num_L_calls +
## Hour + Year, data = AC.DF1Co)
##
## Residuals:
## Min 1Q Median 3Q Max
## -5.2512 -1.5882 -0.2654 1.3150 9.4266
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 103.525252 0.483267 214.220 < 2e-16 ***
## Tot_Knocks 0.016028 0.003569 4.490 1.26e-05 ***
## Num_Herbivory 0.057785 0.025692 2.249 0.0257 *
## Num_L_calls 0.022510 0.029488 0.763 0.4462
## Hour21 0.447799 0.605546 0.739 0.4606
## Hour3 -0.526655 0.603192 -0.873 0.3837
## Hour9 1.565232 0.606584 2.580 0.0107 *
## Year18 3.638921 0.384871 9.455 < 2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 2.65 on 183 degrees of freedom
## Multiple R-squared: 0.4303, Adjusted R-squared: 0.4085
## F-statistic: 19.75 on 7 and 183 DF, p-value: < 2.2e-16
The addition of year and hour took care of variation in total knocks, herbivory and long calls. With this variation accounted for, the effect size of knocks increased to 0.01. The effect size of herbivory also increased dramatically (it was obscured by hour and/or year) to 0.05. The only hour that was significantly different was 9AM and it was 1.56 dB louder than 3PM (our baseline). Year was also significant, 2018 was 3.64 dB louder than 2017.
##
## Call:
## lm(formula = SPL_Midrange ~ Tot_Knocks + Num_Herbivory + Num_L_calls +
## Hour + Year + Tot_Knocks:Hour, data = AC.DF1Co)
##
## Residuals:
## Min 1Q Median 3Q Max
## -5.2885 -1.4974 -0.2355 1.1386 6.7197
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 102.731561 0.486325 211.241 < 2e-16 ***
## Tot_Knocks -0.014904 0.009263 -1.609 0.109381
## Num_Herbivory 0.079026 0.023248 3.399 0.000832 ***
## Num_L_calls 0.022666 0.026201 0.865 0.388131
## Hour21 1.401574 0.582757 2.405 0.017183 *
## Hour3 0.457014 0.584719 0.782 0.435479
## Hour9 2.256687 0.587913 3.838 0.000171 ***
## Year18 3.526664 0.341585 10.324 < 2e-16 ***
## Tot_Knocks:Hour21 0.018812 0.011008 1.709 0.089182 .
## Tot_Knocks:Hour3 0.020202 0.010769 1.876 0.062294 .
## Tot_Knocks:Hour9 0.064360 0.010882 5.914 1.64e-08 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 2.349 on 180 degrees of freedom
## Multiple R-squared: 0.5598, Adjusted R-squared: 0.5353
## F-statistic: 22.89 on 10 and 180 DF, p-value: < 2.2e-16
So with the addition of the interaction term, we see that total knocks itself is no longer significant, this means that its original significance is actually more/better explained by the interaction between knocks and hour. In the interaction we see that the only significant level is between knocks and hour at 9 AM. The effect of this is -0.0149 + 0.0643 = 0.0494, meaning that for every 20 knocks at 9 AM you get an increase of 1 SPL.
The addition of this interaction term also apparently takes care of some variation around 9PM, now allowing us to see that 9PM is 1.4 dB louder than 3PM. Herbivory also now has a larger effect size of 0.079.
##
## Call:
## lm(formula = SPL_Midrange ~ Tot_Knocks + Num_Herbivory + Num_L_calls +
## Hour + Year + Tot_Knocks:Hour + Num_Herbivory:Hour, data = AC.DF1Co)
##
## Residuals:
## Min 1Q Median 3Q Max
## -4.6108 -1.4726 -0.0759 1.1891 6.7452
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 102.627322 0.487568 210.488 < 2e-16 ***
## Tot_Knocks -0.016241 0.009228 -1.760 0.080132 .
## Num_Herbivory 0.094851 0.024414 3.885 0.000144 ***
## Num_L_calls 0.029665 0.026522 1.119 0.264862
## Hour21 1.494018 0.580547 2.573 0.010888 *
## Hour3 -2.377619 2.275854 -1.045 0.297579
## Hour9 1.966747 2.195197 0.896 0.371506
## Year18 3.497448 0.341653 10.237 < 2e-16 ***
## Tot_Knocks:Hour21 0.023357 0.011202 2.085 0.038502 *
## Tot_Knocks:Hour3 0.021779 0.010724 2.031 0.043770 *
## Tot_Knocks:Hour9 0.065626 0.010878 6.033 9.19e-09 ***
## Num_Herbivory:Hour21 -0.140849 0.075533 -1.865 0.063875 .
## Num_Herbivory:Hour3 -1.028214 0.755160 -1.362 0.175060
## Num_Herbivory:Hour9 -0.162766 0.735645 -0.221 0.825148
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 2.334 on 177 degrees of freedom
## Multiple R-squared: 0.5728, Adjusted R-squared: 0.5414
## F-statistic: 18.25 on 13 and 177 DF, p-value: < 2.2e-16
When we add our second interaction effect (Herbivory*Hour), we are saying that the relationship between Herbivory sounds and SPL changes depending on the hour.
We see that the actual effect size of herbivory increases (I don’t know why yet), however, none of the levels of the interaction here are significant. This indicates that hour does not actually change the effect of herbivory on SPL.
In doing this, however, we remove variation between total knocks and hour, and increase effect size. Now we see that the interaction between knocks and hour is significant at all levels. Although, the additive effects are still very small at 9PM (0.0073) and at 3AM (0.00578), effectively making them null.
Finally, we lose 9AM as a significantly different time from 3PM. This is because its differences must be associated with the Herbivory and Hour interaction.
##
## Call:
## lm(formula = SPL_Midrange ~ Tot_Knocks + Num_Herbivory + Num_L_calls +
## Hour + Year + Tot_Knocks:Hour + Num_Herbivory:Hour + Num_L_calls:Hour,
## data = AC.DF1Co)
##
## Residuals:
## Min 1Q Median 3Q Max
## -4.8724 -1.4698 -0.2069 1.1333 6.2390
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 102.495774 0.484169 211.694 < 2e-16 ***
## Tot_Knocks -0.016528 0.009029 -1.831 0.0689 .
## Num_Herbivory 0.095961 0.023900 4.015 8.82e-05 ***
## Num_L_calls -0.069837 0.096339 -0.725 0.4695
## Hour21 1.422829 0.572748 2.484 0.0139 *
## Hour3 -2.604241 2.260053 -1.152 0.2508
## Hour9 2.550893 2.155806 1.183 0.2383
## Year18 3.568114 0.335229 10.644 < 2e-16 ***
## Tot_Knocks:Hour21 0.025312 0.010981 2.305 0.0223 *
## Tot_Knocks:Hour3 0.021447 0.010512 2.040 0.0428 *
## Tot_Knocks:Hour9 0.062264 0.010727 5.804 2.99e-08 ***
## Num_Herbivory:Hour21 -0.164139 0.074319 -2.209 0.0285 *
## Num_Herbivory:Hour3 -1.125353 0.747252 -1.506 0.1339
## Num_Herbivory:Hour9 0.207310 0.732201 0.283 0.7774
## Num_L_calls:Hour21 0.141990 0.100755 1.409 0.1605
## Num_L_calls:Hour3 0.023735 0.126434 0.188 0.8513
## Num_L_calls:Hour9 -0.173272 0.141035 -1.229 0.2209
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 2.283 on 174 degrees of freedom
## Multiple R-squared: 0.5982, Adjusted R-squared: 0.5612
## F-statistic: 16.19 on 16 and 174 DF, p-value: < 2.2e-16
Now we add our final interaction between long calls and hour. This means that we also believe that the relationship between long calls and SPL changes based on the hour we sampled.
We find that this does not seem to be the case, however, in doing this, we have removed some variation that obscured the relationship between herbivory and hour at 9PM. (I worry here that at this step/possibly the last step we are overfitting?) We find that at 9PM, the effect size of Herbivory is actually negative (which doesn’t make logical sense) (-0.0641).
## Linear mixed model fit by REML. t-tests use Satterthwaite's method [
## lmerModLmerTest]
## Formula: SPL_Midrange ~ Tot_Knocks + Num_Herbivory + Num_L_calls + Hour +
## Year + Tot_Knocks:Hour + Num_Herbivory:Hour + Num_L_calls:Hour +
## (1 | Site)
## Data: AC.DF1Co
##
## REML criterion at convergence: 881.4
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -1.96637 -0.64984 -0.04602 0.52505 2.61169
##
## Random effects:
## Groups Name Variance Std.Dev.
## Site (Intercept) 0.9442 0.9717
## Residual 4.5753 2.1390
## Number of obs: 191, groups: Site, 5
##
## Fixed effects:
## Estimate Std. Error df t value Pr(>|t|)
## (Intercept) 102.488770 0.633574 12.859440 161.763 < 2e-16 ***
## Tot_Knocks -0.019041 0.009069 173.992210 -2.100 0.037204 *
## Num_Herbivory 0.089278 0.022911 172.081483 3.897 0.000139 ***
## Num_L_calls 0.021623 0.092279 172.051789 0.234 0.815013
## Hour21 1.316661 0.545335 171.430449 2.414 0.016813 *
## Hour3 -1.934212 2.147264 171.274496 -0.901 0.368970
## Hour9 2.664335 2.029130 170.480181 1.313 0.190933
## Year18 3.654715 0.314726 170.150269 11.612 < 2e-16 ***
## Tot_Knocks:Hour21 0.023039 0.010394 170.954532 2.217 0.027969 *
## Tot_Knocks:Hour3 0.023985 0.010142 172.549443 2.365 0.019149 *
## Tot_Knocks:Hour9 0.058132 0.010404 172.825185 5.587 8.83e-08 ***
## Num_Herbivory:Hour21 -0.137976 0.070233 170.741418 -1.965 0.051089 .
## Num_Herbivory:Hour3 -0.894827 0.707154 170.897624 -1.265 0.207453
## Num_Herbivory:Hour9 0.274713 0.687808 170.202108 0.399 0.690097
## Num_L_calls:Hour21 0.075658 0.095433 171.058180 0.793 0.428998
## Num_L_calls:Hour3 0.018287 0.118594 170.022168 0.154 0.877637
## Num_L_calls:Hour9 -0.274625 0.137256 172.933191 -2.001 0.046976 *
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation matrix not shown by default, as p = 17 > 12.
## Use print(x, correlation=TRUE) or
## vcov(x) if you need it
In adding site as a random effect, we now hope to remove any variation that exists between sites which hasn’t previously been accounted for.
When we do this, we see that variation now reveals changed effect sizes for the interaction between knocks and hour. At 9AM the effect is reduced to 0.039 while at 3AM and 9PM the effect sizes get even closer to zero.
9PM is still significantly different from 3PM (our baseline), 1.3 dB louder than it.
Lastly, we now see a significant interaction between long calls and 9AM, however it is negative (-0.2546), which doesn’t really make sense.
## Linear mixed model fit by REML. t-tests use Satterthwaite's method [
## lmerModLmerTest]
## Formula: SPL_Midrange ~ Tot_Knocks + Num_Herbivory + Num_L_calls + Hour +
## Year + Tot_Knocks:Hour + (1 | Site)
## Data: AC.DF1Co
##
## REML criterion at convergence: 887.3
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -1.85688 -0.71841 -0.05929 0.63238 2.80744
##
## Random effects:
## Groups Name Variance Std.Dev.
## Site (Intercept) 1.044 1.022
## Residual 4.830 2.198
## Number of obs: 191, groups: Site, 5
##
## Fixed effects:
## Estimate Std. Error df t value Pr(>|t|)
## (Intercept) 102.659592 0.649841 11.847958 157.976 < 2e-16 ***
## Tot_Knocks -0.018835 0.009275 179.998737 -2.031 0.04375 *
## Num_Herbivory 0.071178 0.022191 177.675485 3.208 0.00159 **
## Num_L_calls 0.061210 0.025749 179.604479 2.377 0.01850 *
## Hour21 1.305358 0.553736 177.297508 2.357 0.01950 *
## Hour3 0.500766 0.553766 177.119142 0.904 0.36707
## Hour9 2.346686 0.556795 177.137895 4.215 3.98e-05 ***
## Year18 3.606956 0.320083 176.059912 11.269 < 2e-16 ***
## Tot_Knocks:Hour21 0.018050 0.010375 176.665440 1.740 0.08366 .
## Tot_Knocks:Hour3 0.022464 0.010360 178.365014 2.168 0.03145 *
## Tot_Knocks:Hour9 0.061206 0.010508 178.518970 5.825 2.61e-08 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation of Fixed Effects:
## (Intr) Tt_Knc Nm_Hrb Nm_L_c Hour21 Hour3 Hour9 Year18 T_K:H2
## Tot_Knocks 0.379
## Num_Herbvry -0.272 -0.193
## Num_L_calls 0.030 -0.019 -0.101
## Hour21 -0.525 -0.444 0.340 -0.234
## Hour3 -0.546 -0.466 0.430 -0.035 0.642
## Hour9 -0.538 -0.463 0.421 0.036 0.620 0.673
## Year18 -0.275 -0.023 0.003 0.058 0.030 0.036 0.031
## Tt_Knck:H21 -0.296 -0.808 0.123 0.049 0.287 0.362 0.363 0.009
## Tt_Kncks:H3 -0.329 -0.856 0.191 0.016 0.386 0.375 0.405 0.015 0.712
## Tt_Kncks:H9 -0.316 -0.845 0.189 0.002 0.383 0.401 0.365 -0.021 0.706
## T_K:H3
## Tot_Knocks
## Num_Herbvry
## Num_L_calls
## Hour21
## Hour3
## Hour9
## Year18
## Tt_Knck:H21
## Tt_Kncks:H3
## Tt_Kncks:H9 0.739
Here I am reducing our model to only include interactions I think are biologically relevant (knocks*hour) and our random effect. Part of me thinks that this is the most effective model for explaining our system and that the previous model might be overfitting with all of the interaction terms.
Plots for use in paper
This first plot uses ALL data (2017 & 2018) and all sites
It seems super crowded, so maybe we only need to display a piece of it?
Here is a plot using only one site (Site 5)
This plot uses ALL data (2017 & 2018) and all sites/hours
Way too much going on.
Here is a plot that only uses 9am
This shows that the pattern exists at pretty much all sites except for 5. I think this is the plot to go with.
That looks nice!
Hey! So does that! Is it not OK for us to use site 5 here because it was the one site that didn’t seem to have a strong relationship at 9 AM?
## [1] "21:00" "21:10" "21:20" "21:30" "21:40" "21:50" "22:00" "22:10"
## [9] "22:20" "22:30" "22:40" "22:50" "23:00" "23:10" "23:20" "23:30"
## [17] "23:40" "23:50" "0:00" "0:10" "0:20" "0:30" "0:40" "0:50"
## [25] "1:00" "1:10" "1:20" "1:30" "1:40" "1:50" "2:00" "2:10"
## [33] "2:20" "2:30" "2:40" "2:50" "3:00" "3:10" "3:20" "3:30"
## [41] "3:40" "3:50" "4:00" "4:10" "4:20" "4:30" "4:40" "4:50"
## [49] "5:00" "5:10" "5:20" "5:30" "5:40" "5:50" "6:00" "6:10"
## [57] "6:20" "6:30" "6:40" "6:50" "7:00" "7:10" "7:20" "7:30"
## [65] "7:40" "7:50" "8:00" "8:10" "8:20" "8:30" "8:40" "8:50"
## [73] "9:00" "9:10" "9:20" "9:30" "9:40" "9:50" "10:00" "10:10"
## [81] "10:20" "10:30" "10:40" "10:50" "11:00" "11:10" "11:20" "11:30"
## [89] "11:40" "11:50" "12:00" "12:10" "12:20" "12:30" "12:40" "12:50"
## [97] "13:00" "13:10" "13:20" "13:30" "13:40" "13:50" "14:00" "14:10"
## [105] "14:20" "14:30" "14:40" "14:50" "15:00" "15:10" "15:20" "15:30"
## [113] "15:40" "15:50" "16:00" "16:10" "16:20" "16:30" "16:40" "16:50"
## [121] "17:00" "17:10" "17:20" "17:30" "17:40" "17:50" "18:00" "18:10"
## [129] "18:20" "18:30" "18:40" "18:50" "19:00" "19:10" "19:20" "19:30"
## [137] "19:40" "19:50" "20:00" "20:10" "20:20" "20:30" "20:40" "20:50"
## [145] "11:45" "12:15" "12:45" "13:15" "13:45" "14:15" "14:45" "15:15"
## [153] "15:45" "16:15" "16:45" "17:15" "17:45" "18:15" "18:45" "19:15"
## [161] "19:45" "20:15" "20:45" "21:15" "21:45" "22:15" "22:45" "23:15"
## [169] "23:45" "0:15" "0:45" "1:15" "1:45" "2:15" "2:45" "3:15"
## [177] "3:45" "4:15" "4:45" "5:15" "5:45" "6:15" "6:45" "7:15"
## [185] "7:45" "8:15" "8:45" "9:15" "9:45" "10:15" "10:45" "11:15"
## [1] "21:00" "21:10" "21:20" "21:30" "21:40" "21:50" "22:00" "22:10"
## [9] "22:20" "22:30" "22:40" "22:50" "23:00" "23:10" "23:20" "23:30"
## [17] "23:40" "23:50" "0:00" "0:10" "0:20" "0:30" "0:40" "0:50"
## [25] "1:00" "1:10" "1:20" "1:30" "1:40" "1:50" "2:00" "2:10"
## [33] "2:20" "2:30" "2:40" "2:50" "3:00" "3:10" "3:20" "3:30"
## [41] "3:40" "3:50" "4:00" "4:10" "4:20" "4:30" "4:40" "4:50"
## [49] "5:00" "5:10" "5:20" "5:30" "5:40" "5:50" "6:00" "6:10"
## [57] "6:20" "6:30" "6:40" "6:50" "7:00" "7:10" "7:20" "7:30"
## [65] "7:40" "7:50" "8:00" "8:10" "8:20" "8:30" "8:40" "8:50"
## [73] "9:00" "9:10" "9:20" "9:30" "9:40" "9:50" "10:00" "10:10"
## [81] "10:20" "10:30" "10:40" "10:50" "11:00" "11:10" "11:20" "11:30"
## [89] "11:40" "11:50" "12:00" "12:10" "12:20" "12:30" "12:40" "12:50"
## [97] "13:00" "13:10" "13:20" "13:30" "13:40" "13:50" "14:00" "14:10"
## [105] "14:20" "14:30" "14:40" "14:50" "15:00" "15:10" "15:20" "15:30"
## [113] "15:40" "15:50" "16:00" "16:10" "16:20" "16:30" "16:40" "16:50"
## [121] "17:00" "17:10" "17:20" "17:30" "17:40" "17:50" "18:00" "18:10"
## [129] "18:20" "18:30" "18:40" "18:50" "19:00" "19:10" "19:20" "19:30"
## [137] "19:40" "19:50" "20:00" "20:10" "20:20" "20:30" "20:40" "20:50"
## [145] "11:45" "12:15" "12:45" "13:15" "13:45" "14:15" "14:45" "15:15"
## [153] "15:45" "16:15" "16:45" "17:15" "17:45" "18:15" "18:45" "19:15"
## [161] "19:45" "20:15" "20:45" "21:15" "21:45" "22:15" "22:45" "23:15"
## [169] "23:45" "0:15" "0:45" "1:15" "1:45" "2:15" "2:45" "3:15"
## [177] "3:45" "4:15" "4:45" "5:15" "5:45" "6:15" "6:45" "7:15"
## [185] "7:45" "8:15" "8:45" "9:15" "9:45" "10:15" "10:45" "11:15"
## [1] "21:00" "21:10" "21:20" "21:30" "21:40" "21:50" "22:00" "22:10"
## [9] "22:20" "22:30" "22:40" "22:50" "23:00" "23:10" "23:20" "23:30"
## [17] "23:40" "23:50" "0:00" "0:10" "0:20" "0:30" "0:40" "0:50"
## [25] "1:00" "1:10" "1:20" "1:30" "1:40" "1:50" "2:00" "2:10"
## [33] "2:20" "2:30" "2:40" "2:50" "3:00" "3:10" "3:20" "3:30"
## [41] "3:40" "3:50" "4:00" "4:10" "4:20" "4:30" "4:40" "4:50"
## [49] "5:00" "5:10" "5:20" "5:30" "5:40" "5:50" "6:00" "6:10"
## [57] "6:20" "6:30" "6:40" "6:50" "7:00" "7:10" "7:20" "7:30"
## [65] "7:40" "7:50" "8:00" "8:10" "8:20" "8:30" "8:40" "8:50"
## [73] "9:00" "9:10" "9:20" "9:30" "9:40" "9:50" "10:00" "10:10"
## [81] "10:20" "10:30" "10:40" "10:50" "11:00" "11:10" "11:20" "11:30"
## [89] "11:40" "11:50" "12:00" "12:10" "12:20" "12:30" "12:40" "12:50"
## [97] "13:00" "13:10" "13:20" "13:30" "13:40" "13:50" "14:00" "14:10"
## [105] "14:20" "14:30" "14:40" "14:50" "15:00" "15:10" "15:20" "15:30"
## [113] "15:40" "15:50" "16:00" "16:10" "16:20" "16:30" "16:40" "16:50"
## [121] "17:00" "17:10" "17:20" "17:30" "17:40" "17:50" "18:00" "18:10"
## [129] "18:20" "18:30" "18:40" "18:50" "19:00" "19:10" "19:20" "19:30"
## [137] "19:40" "19:50" "20:00" "20:10" "20:20" "20:30" "20:40" "20:50"
## [145] "11:45" "12:15" "12:45" "13:15" "13:45" "14:15" "14:45" "15:15"
## [153] "15:45" "16:15" "16:45" "17:15" "17:45" "18:15" "18:45" "19:15"
## [161] "19:45" "20:15" "20:45" "21:15" "21:45" "22:15" "22:45" "23:15"
## [169] "23:45" "0:15" "0:45" "1:15" "1:45" "2:15" "2:45" "3:15"
## [177] "3:45" "4:15" "4:45" "5:15" "5:45" "6:15" "6:45" "7:15"
## [185] "7:45" "8:15" "8:45" "9:15" "9:45" "10:15" "10:45" "11:15"